W3C home > Mailing lists > Public > www-rdf-logic@w3.org > May 2001

Re: What do the ontologists want

From: pat hayes <phayes@ai.uwf.edu>
Date: Fri, 18 May 2001 15:01:18 -0500
Message-Id: <v0421012eb72b2b20199c@[]>
To: Sandro Hawke <sandro@w3.org>
Cc: www-rdf-logic@w3.org
>pay hayes:
> >                                                        The way I
> > understand it, reification means treating the symbols as objects.
>OH!!!!!!!!!!!!!!     Duh!  I get it now!!
>In general, I think reification means to treat something as a
>first-class object, to put it in your domain of discourse.

Ah, OK, let's go with that as a definition.

>So reification of lists, etc, is just building data structures, which
>is fine.  Reification of RDF Triples is fine, too, you say.

Well...sorry to be niggling, but there is a difference between 
*using* datastructures, on the one hand, and *quantifying over* them, 
on the other. You might want to do both, of course, but that seems a 
little advanced at this stage.

Look, suppose I implement a Prolog interpeter in LISP. The Prolog 
assertions are coded as LISP datastructures - lists - but that doesnt 
automatically put lists into the Prolog domain of discourse. The 
Prolog variables might not range over lists at all: and even if they 
do - say, if I write a simple theory of lists in clause form - then 
the lists in the Prolog domain won't be the same lists that the LISP 
is talking about (unless I do some very clever behind-the-scenes 
tricks in the interpreter to keep the various encodings in line with 
each other, which would kind of do an end-run around the usual Prolog 
semantics.) The picture I have of using RDF triples is like this kind 
of use of LISP to implement some other language. This really doesnt 
seem like reification even in your sense. Well, I guess it could be 
for the RDF, but then as soon as it gets to be, that is where the RDF 
and the extended language part company, just as the Prolog has to 
make a clear disitnction between the lists used to implement its 
syntax, and the lists that it might be talking about. So there would 
be two domains of discourse: in the RDF domain, thre are 
datastructures; in the domain of the language described by the 
datastructures, there would be something else (exactly what would 
depend on the particular ontology.)

Having a language able to talk baout the structures which implement 
its *own* syntax is logically(semantically) very hairy. LISP does 
that, of course, and it is a terrific prototyping tool for that very 
reason; but LISP is not an assertional language, and giving a model 
theory to LISP was a very, very hard problem (Dana Scott did it, but 
it was a real tricky business and is almost unusable in practice.) So 
I guess I would just kind of recommend that RDF not go that road, 
since it seems like the opposite end of a spectrum from what RDF 
wants to be, ie a kind of simple, easy, usable model-T kind of 

>It's reificiation of SYMBOLS which is so worrisome to you.   That I
>can understand!

Reification of expressions, yes. With my background, the term 
'reification' is pretty much restricted to this case.

>I've played around a little with redefining the basic RDF model to be
>triples of a Relation, a String, and a String, and letting the
>Relation take care of mapping the String to its denotation.  Which
>seems like it's heading in that direction.  To actually put the
>relation identifier itself in the domain of discourse....  hrm...
>Right now I don't actually see the danger there, but it DOES scare me,
>and I have no particular desire to do it.  And I don't think I've ever
>seen anyone in the RDF community try or want to do it, except as a
>bizarre twisted experiment.  Or maybe I've been misunderstanding,
>interpreting things in the way which seemed most reasonable and safe
>to me.
>The closest I think I've seen is Tim B-L's
>   <> log:forAll :x
>which probably means "everything asserted in this document is also
>asserted with every possible symbol subsituted for the symbol 'x'
>except in this assertion itself".  That's symbol reification

No, I think that is just universal quantification, which doesn't 
involve reification. (To describe its truth-conditions does, of 
course, but then to describe any truth-conditions does; but that is 
in the metalanguage.)

>, and
>fairly worrisome.  It could, alternatively mean "everything asserted
>in this document is also asserted with all uses of the object denoted
>by 'x' replaced by any other object, except in this statement," which
>is still odd, but does not involve symbol reification.
>So: can we try to avoid using the word "reification" unless we're
>saying what we're considering reifying?   Does that solve the problem?

That would certianly clear the air, but I think it doesnt solve the 
basic problem. (See previous messages.)


IHMC					(850)434 8903   home
40 South Alcaniz St.			(850)202 4416   office
Pensacola,  FL 32501			(850)202 4440   fax
Received on Friday, 18 May 2001 16:01:22 UTC

This archive was generated by hypermail 2.4.0 : Friday, 17 January 2020 22:45:37 UTC